2. Methodology
2.10 Data Analysis
were shared with some of the members of the executive teams from each of the two companies, as well as the Chief Executive Officer from one of the two companies, before being reported publicly.
2.9.4 Archival Data
Materials from governmental publications, press reports, and online documents were examined.
The JSE Socially Responsible Investment (SRI) Index and the Carbon Disclosure Project (CDP) for the years 2008-2010 were reviewed and non-financial reports of the two companies were downloaded from the corporations‘ websites. One of the companies provided a hard copy of their Sustainability and Climate Change Strategy as well as a copy of their Environmental Sustainability Action Plan. Following on the methodology followed by Ihlen (2009), a keyword/phrase search for ―climate change‖, ―global warming‖, ―greenhouse gas‖, ―carbon/ CO2
emissions‖, ―COP 15/17‖, ―UNFCC‖, ―IPCC‖, was conducted. The reports were also read in their original context and categories and stances on climate change were synthesised. The five typical environmental stances suggested by Bullis and Ie (2007) were adopted in this study, namely: compliance (reacting to pressure), openness (information sharing), integration (attempting to realise positive gains), collaboration (partnering with external stakeholders) and sustainability (implementing an ethical, ecological, and systems-based approach that does not place the entity‘s financial interests first). The findings from this exercise were merged (compared and contrasted) with the results from the primary data collection process, thereby completing the data collection and analysis triangulation process.
Primary and secondary data collection and analysis was used in an interactive and sequential manner. Secondary data and the literature review informed the preparation and conduct of the interviews, while findings from the one-on-one interviews led to the refining of the literature and a further literature search. Because the interviewees were business executives charged with the crafting and subsequent implementation of a climate change strategy, their statements were taken to be authoritative. Literature was still used to triangulate the statements however.
2.10.1 Qualitative Content Analysis
Content analysis is the qualitative research technique used in this thesis. There are three distinct approaches to qualitative content analysis namely: conventional, directed, or summative (Hsieh & Shannon, 2005). All three approaches are used to interpret meaning from the content of text data thus, adhere to the naturalistic paradigm. Qualitative content analysis is an analytical approach of empirical, methodologically-controlled characterisation of language as communication, with attention to its content and context (Krippendorff, 2004) in order to understand meanings, themes and patterns in the particular text (Wheelock et al., 2000; Hsieh
& Shannon, 2005) and knowledge of the phenomenon under study (Weber, 1990). Holsti (1969:14) offered a broader definition of content analysis as, "any technique for making inferences by objectively and systematically identifying specified characteristics of messages".
In this study, Heish and Shannon‘s (2005: 1278) definition of qualitative content analysis was used, which is, ―the subjective interpretation of the content of text data through the systematic classification process of coding and identifying themes or patterns‖.
The approaches to coding scheme generation, origins of the codes and threats to trustworthiness differentiate the different approaches (Hsieh & Shannon, 2005; Krippendorff, 2004) to qualitative content analysis. Hsieh and Shannon (2005) advanced three main approaches to interpret meaning from content data, namely: conventional content analysis, where coding categories are derived directly from the text data; directed content analysis, where initial coding is guided by some theory or relevant research findings; and summative content analysis, which involves counting and comparisons of key words followed by the interpretation of the underlying context.
Directed content analysis, also referred to as inductive category content analysis and summative (latent) content analysis, were the primary qualitative content analysis approaches used in this thesis. Simon‘s bounded rationality concept was the primary theory under study, with the study objective being to practically validate and/or extend this theoretical framework in behavioural strategy applications. The motivations and drivers of climate change response and the categories of initiatives undertaken by businesses are the other two classes of content which were also analysed using qualitative content analysis.
Figure 2-2 Content Analysis Process Flow
Source: Krippendorff, 2004
From the process flow diagram shown in Figure 2-2, bounded rationality theory and climate change response strategy theory were used to develop the initial coding schemes. Through the
Research Objective, Research Questions
Determination of category definition Defined in Literature Review section:
Motivations/drivers
Bounds of rationality
Climate change response initiatives
Develop categories and coding scheme
Test coding scheme and define consistency Prepare data
Data was transcribed verbatim – excluding observations during the interview e.g. sounds, audible mannerisms, etc
Define Unit of Analysis categories Single word, phrases, sentence, paragraph
Revise categories after 10-20% of the material Code all the text
Formative check of reliability Summative check of reliability Interpret and draw conclusions from coded data
loopback process shown above, additional codes were developed and the initial coding scheme was revised and refined as data analysis progressed.
2.10.2 Directed (inductive category) Content Analysis
With a directed approach, analysis starts off with a theory as the basis for identifying key themes, concepts and categories and acts as a guide for the generation of initial codes (Boyatzis, 1998). The main idea of the procedure is to formulate a criterion of definitions (derived from theoretical background and research questions). Following a step-by-step work through content material, descriptive labels and tentative labels are given to distinct occurrences of phenomena, gradually moving down to lower-level categories and integrating these into higher-level categories (Hsieh & Shannon, 2005) as necessary. Using an iterative process, the categories are revised, reworked and eventually reduced to main categories which are checked for reliability. The integration of the data allows for insightful categorization of information, promoting the emergence and generation of new theory (Mayring, 2000).
The theory of bounded rationality, technology adoption, the ‗no regret paradox‘ and ‗energy- efficiency gap‘ theories were used to provide variables of interest and the relationships among the variables (Hsieh & Shannon, 2005). Porter and Reinhardt‘s concepts of corporate climate change response initiatives were also used as a second set of coding categories. Having identified the key concepts and variables, operational definitions for each category were determined using the literature review. Professional colleagues, an industrial psychologist and climate change response specialists agreed to the coding categories that were applied to the data. Revisions to the categories were made, up to the point where exclusivity and exhaustiveness were maximised (Boyatzis, 1998; Weber, 1990).
2.10.3 Summative (latent) Content Analysis
Summative analysis is particularly useful for texts that are complex or cover sensitive topics. It works best with diverse data, snippets of text, or long, meandering tracts and enables data that do not fit a mould to be considered, offering more analysis flexibility (Rapport, 2010).
Summative analysis has the potential to generate a range of insights and reflections and can provide a careful approach to sensitive or emotional topics (Sparkes, 2002) by allowing the
researcher to become aware of ambiguous aspects of text. It preserves the quality of the speaker‘s voice irrespective of the mode of presentation and considers issues beyond a researcher‘s own knowledge (Rapport, 2010). The analysis technique‘ greatest strengths are its versatility and inclusivity where everyone‘s views matter. While end-result of analysis is reduction (Miles & Huberman, 1994), summative analysis minimises the reductive effect by considering the importance of text as a whole and its impact on the speaker and audience (Rapport, 2010).
In order to build the climate change response continuum, a summative approach to content analysis was used, where the starting point was the identification and quantification of certain words and phrases in the text with the purpose of understanding the contextual use of the words. The data analysis started with computer-assisted searches (using NUD*IST) for occurrences of the terms of potential interest such as ―water utilisation‖, ―energy efficiency‖,
―waste‖, ―awareness‖, etc. (Annexure 4) for the coded scripts. This was then followed by a Key Word in Context (KWIC) search to test for the consistency of usage of the words (Weber, 1990). Word frequency counts for each of the climate change actions – related words or phrases were calculated and compared per speaker. Next, alternative terms or expressions used instead of ―waste‖, energy efficiency‖, ―water utilisation‖, etc., were identified. Occurrences of these terms were also counted as a total number and for each alternative term. This quantification was an attempt to explore usage (manifest content analysis) and prevalence, as well as the extent to which these climate change initiatives were a part of the company‘s current climate change response agenda or under consideration.
The transcripts were then reviewed carefully, highlighting all text that on first impression appeared to describe bounded rationality in strategic decision making. Data that could not be coded into the identified and agreed coding categories were re-examined to search for new and emerging categories or sub-categories of existing codes. The coded data was analysed to establish the extent to which the research findings offered supporting or discomfiting evidence of strategic decision making biases and bounds on rationality. The extent to which findings that were supportive of bounded rationality theory compared with rational economic utility maximisation theory were examined. Incidences of the codes representing bounds on rationality are presented, with examples from the interview scripts and descriptive evidence, where available, also being provided.
2.10.4 The Use of Categories
Categorisation or coding of data was undertaken to facilitate an understanding, retrieval and comparison of the information within and between case studies. A priori categories of analysis were determined based on Table 2-1, but most QDA packages allow for the emergence of new categories based on the analysis. Both options were used in order to allow for comparability as well as to identify outliers requiring further analysis. Descriptive, analytical and interpretative categories were then explored in the study. Coding indicating the frequency (number of counts) of themes and categories of climate change response initiatives, while binary counts (presence or absence) of certain categories were also used in the analysis (Jehn & Jonsen, 2010). The quantitised frequencies indicated particularly prominent initiatives, drivers or irrational behaviours in organisational strategic decision making processes (Onwuegbuzie & Teddlie, 2003). The quantification of qualitative data enabled an inter-company comparison, as well as a comparison with the quantitative data collected within an organisation. Pattern analysis, the method of analysis indicated for multiple case studies, entailed identifying a pattern of categories and variables (Seekamp et al., 2010, Dures et al., 2010) from the literature and the pilot case, and then assessing patterns observed in each of the two cases against that predicted pattern. The binary form of quantitising determined the presence or absence of each behavioural category for each participant. Individual responses from the one-on-one interviews were converted into a series of coded response categories that were in turn quantified as binary variables 0 or 1 based on the absence or presence of each coded response.
Because quantitising has the effect of reducing everything to a single-dimension, causing considerable loss of context and meaning (Seekamp, et al., 2010, Rocco, et al., 2003), interpretation of these categories was based on both the quantitisied and the raw transcripts.
Where evidence from different sources was inconsistent or contradictory, reasons for the discrepancies were sought.
2.10.5 Interpretive Content Analysis
Text analysis was used for construct development and construct measurement in a variety of organisational contexts (Jehn & Jonsen, 2010; Seekamp et al., 2010) to facilitate data
exploration. To identify and measure the irrational behavioural constructs in this study, interpretive content analysis (Seekamp, 2010) was employed to assess the types of constructs and social influences reflected in the statements made by participants in the one-on-one and group deliberative processes, and to develop coding categories for the behavioural constructs.
The focus on the interpretive content analysis was on understanding the meaning of data in context to catch salient as well as subtle behaviours under study. A trained psychologist was utilised for the process to reduce researcher biases, such as reactivity, selective information processing and possible misinterpretation.
2.10.6 Examination of Other Data Sources
Transcriptions of field notes and notes taken during observations were analysed for instances of the irrational behavioural constructs.